The invention relates to a method for selectively restoring a sound signal delivered by an electric source in several parallel mounted transducers, each transducer being specialized for restoring a frequency range of the spectrum of the signal. The method consists more precisely in permanently fractionating the sound signal into frequencies so as to assign each component to the transducer optimized for the corresponding frequency range, without using any additional selection circuit such as filtering circuits or others.
The invention also relates to all the associated sets of transducers for implementing the method.